Some things in IT never change; you can never have enough memory, speed, or storage—except when the rules do in fact change in time, and that extra capacity causes a problem in managing your virtual machines (VMs).
That surprising enlightenment came from Adrian Moir, senior consultant and evangelist at Quest Software, when asked about the latest and greatest VM performance tips.
"One of the things we find a lot, and it's kind of a hot topic at the moment with people wanting to move to the cloud, [VM tuning] is kind of dependent on the infrastructure itself," Moir said. "They've taken the physical model of what a server needs and applied that to a virtual machine, and what happens is they over-provision. You're going to be paying an awful lot of money for things you don't use."
SEE: Virtualization policy (Tech Pro Research)
Companies are better off investing in their networks, Moir asserted. Leave it up to software—of course, that's what Quest makes—to inform your system administrators when extra server capacity or VM adjustments are needed, and in some cases for the software to make those changes on its own accord. Quest's software and similar products can operate by straightforward numeric thresholds or, in more advanced cases, by system behavior analysis.
Moir said tough mistakes when setting up and maintaining virtual machines include:
- Misconfigured VM reservations, such as taking resources away from critical systems;
- losing track of obsolete VMs; and
- virtualizing systems such as transactional databases, which are best left to their own bare-metal servers.
Steven Dreher, director of solutions architecture at Green House Data, says a way to avoid CPU contention is by using a greater number of smaller virtual machines rather than a small number of large ones. Dreher also suggests ensuring your VMs aren't hitting their memory limit, using agent-less antivirus tools because those are more efficient on a cloud, and deleting unneeded snapshots as they're resource hogs.
HiveIO's Kevin McNamara, co-founder/CTO, observed that graphics are an overlooked part of virtual machine optimization. "Many enterprises and OEMs are investing significant money purchasing graphic acceleration cards to offset CPU demand and accommodate for sophisticated 3D accelerated application workloads," he said. "Adding dedicated graphic cards reduces end-point CPU demand and smooths audio/visual experience for VDI workloads. Additionally, enabling CPU pass-through removes transients from server-side workloads."
Sophie Miles, a startup entrepreneur in Europe, shared tips of her own:
- Use fixed disk sizes rather than dynamic ones to make a VM work faster.
- It's more important to run a virtual machine on fast storage than large storage.
- Don't neglect updating a VM's supporting software packages.
- Take the time to try different levels of CPU core assignments for the best balance of speed vs. resources.
- Suspend a VM rather than shutting it down so it starts faster next time while still not using hardware resources.
- How to migrate your current virtual machines to an Antsle cloud server (TechRepublic)
- Microsoft Hyper-V: The smart person's guide (TechRepublic)
- Hyperconverged infrastructure: A cheat sheet (TechRepublic)
- Photos: 20 innovative data centers that give us a glimpse into the future of computing (TechRepublic)
- Quick glossary: Virtualization (Tech Pro Research)
Evan became a technology reporter during the dot-com boom of the late 1990s. He published a book, "Abacus to smartphone: The evolution of mobile and portable computers" in 2015 and is executive director of Vintage Computer Federation, a 501(c)3 non-profit organization. His vices include running and Springsteen.